Rewiring for AI: From ambition to advantage

Added: 2026-05-07

[00:00:01]I think we're less than interested in the number of successful MVPs. We're more interested in how many of these end-to-end workflows have fully scaled across the entire footprint of a company where it's relevant. That's McKinsey partner Rob Levin talking about why leaders should focus on company-wide processes instead of [music] isolated pilots when it comes to AI transformation. He joins me and Kate Smaje, McKinsey's global leader for tech and AI, to discuss their new edition of their best-selling book Rewired, which is all about how companies can outperform in the age of AI.

[00:00:39]>> [music] >> This is the McKinsey Podcast where we help you make sense out of the world's toughest business challenges.

[00:00:46]I'm your host for today, Lucia Rahilly.

[00:00:49]>> [music] >> Okay, let's get rolling. First, congratulations to both of you on the book. Again, the first edition of Rewired was a best-seller and offered readers a really thoughtful framework for how companies could empower themselves with tech and AI. Why a second edition? And Rob, let's start with you. We like to say that that you know, the world didn't give us a choice since um most of the world saw ChatGPT for the first time three and a half years ago. You know, technology has changed as AI moved from machine learning into generative AI, agentic AI, this new capability to actually automate uh end-to-end workflows. And probably the most fundamental disruption is, you know, the disruption of GenAI to software development, right? Code writing, code writing, code. So, so much has changed and we really wanted to go look back at the framework we had established three years ago and sort of say, you know, does this still work?

[00:01:47]Does this recipe uh for established companies to sort of um organize, align, build, adopt, and scale AI, does it work?

[00:01:57]Um as we kept coming back to this this quote actually um you know, which is by the Greek poet Archilochus, you know, we don't let rise to the level of our expectations, we fall to the level of our training. And that feels pretty true about this moment in AI and kind of the core thesis of Rewired. We wanted to check this time around as generative AI and energetic AI has started to become such an important capability, um do the companies these capabilities continue to do well? And that was probably the core learning for us in writing it was the companies that have built these capabilities in sort of AI 1.0 have succeeded far, you know, far more than any of the other um companies who hadn't as we go up into kind of AI 2.0.

[00:02:38]Freeport-McMoRan is a great example of that. Um we there was a great story about Freeport um you know, in the first book based on building a digital twin of the entire copper concentrator, um creating all those efficiencies end-to-end across that process of taking copper out of the rock, and they you know, have were incredibly successful um in driving value.

[00:02:59]They then turned their attention to generative AI, approached another area of the business which is leaching, the final process chemical process to get the uh ore out of the rock, and created a whole bunch of additional value. This new world and it's changing every day and obviously leaders need to be developing new kinds of flexibility and as you put it in Rewired, a second muscle to be able to adapt continuously to these changes.

[00:03:27]Always think of you know, when when they put you on that destabilizing Bosu in Pilates and expect you just carry on exercising. So, on that notion of transformation and continuously adapting to what otherwise could be a a destabilizing change, Kate, talk to us about the scale of what needs to happen and what's at stake here if we accept that AI-enabled transformation is a must. Yeah, I think your analogy of of the destabilizer is is actually a good one because it is hard, right? But one of the things I was excited about with the second edition is just getting a lot grittier about well, what is that value capture? Because so much of this comes down to, you know, I've got everywhere AI everywhere in my organization except for the bottom line. So, what we wanted to do was get granular about well, where do we see that impact? And I want to share just just sort of three three numbers with you around this. And this is looking specifically at a cohort within the research set of 20 companies that are really doing this well. They've really applied the Rewired framework soup to nuts, every part of the six, you know, boxes and so on. And we see three things here. Number one is among those 20, they have on average an EBITDA uplift of 20%. So, to your point about whether or not this is worth it, I think a 20% EBITDA uplift is worth it. And that is just the average, right? We see higher, we see slightly lower.

[00:04:49]Um but secondly, on average it's 1 to 2 years to become cash accretive. That's probably another 1 to 2 to to really kind of fully, you know, start to to to make the money flow. But it but it isn't forever, right? You can actually do this, you know, relatively quickly. And what's really interesting within that is you know, the for most of them, 2/3 of the of the co- cohort, they were able to do this with three or fewer domains. So, again, they're not papering AI everywhere across their organization.

[00:05:19]They're being incredibly focused in where they point the the resource, the money, you know, etc. And then another number I'll give you, for every dollar that they are spending of investment, they're getting $3 back on average. And again, in the grand scheme of of returns that you can get, that's not too shabby.

[00:05:36]So, I think to your point, do we see the value is there? Yes, we absolutely do.

[00:05:40]But that's the cohort of the 20 really really applying this. Thanks, Kate.

[00:05:44]Okay, so those numbers make the case for why why transformation matters and the value at stake is pretty compelling.

[00:05:52]Let's turn to what I imagine is the hard part, which is the how. We know that most transformations fail. What separates the companies that get there from the ones that fall short? Rob, accepting that the two of you have just co-authored an entire book designed to answer this question, give us at least a high level on the framework for answering it. That's the set of capabilities.

[00:06:16]Aligning on a business-led roadmap that does move the needle on the business, focusing on a few domains versus peanut butter spreading use cases across the board.

[00:06:28]Um, you know, a set of enabling capabilities around talent, operating model, the way we work, which is building technology, but also the control functions and the general decision process of a company.

[00:06:40]Um, everything about the tech stack, um, data, which is of course the lifeblood of AI, and how do we make that, um, uh, high quality and constantly usable.

[00:06:50]And then this focus on adoption and scaling. What's new now in the agentic context? A few areas.

[00:06:55]Um, the first in strategy is, you know, end-to-end workflows. We'd always sort of had this concept that thinking about the broader change versus individual use cases had value. That's just become incredibly clear with agentic AI and this ability to actually automate the majority of a tasks of tasks in an end-to-end workflow.

[00:07:13]Um, reimagining that end-to-end versus just sort of deciding on an MVP use case in a workflow is clearly the path to success.

[00:07:22]You know, in the last book we talked a lot about the the challenge of getting density of scarce technology talent and sort of keeping them happy and at top of their craft. This time around we know that like we have to think about the entire workforce because over time agentic AI is coming for, you know, all white-collar workflows and and physical AI will will come for others. Um, And then I think in technology, you know, the this 20x offer development productivity, I mean this incredible fundamental disruption of writing code with the likes of you know, cloud code released in January. It's really collapsing this model of a product owner who knows the definition of what good looks like and a full stack engineer who can you know, work with cloud code, debug it and work it into the architecture. It's never been more complex we would assert to solution technology than this moment with every vendor at every layer of the stack laying their claim to be the center of your AI gravity and many of those capabilities coming with a high level of ongoing opex in terms of cost. So being very thoughtful about this because it's it's easy to think about point solutions or agents in core platforms in every function of the business. But when you step back from that, that's a brittle architecture.

[00:08:37]It's maybe an inefficient architecture.

[00:08:39]Maybe a less secure architecture. And then just lastly adoption. Taking a workflow and reinventing it, that means you know, clean sheet of paper, reimagine the process, take all the roles down to tasks, automate many away, reconstruct them into new roles, train everybody. And so we would say like this is the job to be done of the next several years again and again and again.

[00:09:00]Yeah, it's huge. Okay, so if I'm understanding and synthesizing correctly, the six capabilities here constitute the foundation, but the bar for what good looks like and what's required inside each one has significantly shifted. Let's look a little more closely at that picture.

[00:09:19]Kate, give us a little color on companies that have rewired successfully and what they're doing. One of the biggest differentiators when you really boil these six capabilities together is they're able to operate at a different metabolic rate, right? So their their time, the and almost from insight to decision, from decision to action starts to look really different. It's about outcompeting. And to outcompete, you've got to deploy these capabilities and actually move faster than the than the the the peer next to you. And you know, one of the examples we we talk about in the in the book and tell the story of for a second time is is DBS.

[00:09:57]It's only through three, four years of really hard foundational investments on the part of DBS that when GenAI comes around and and Agility comes around that they're able to move really, really fast. And that's why you start to see in there, you know, on record publicly for this, you know, around a billion of Singaporean dollars of of tangible, verifiable benefits from AI. So, it's both a speed of operation that feels genuinely different, but also the compounding value of these capabilities over time in that you're running faster and faster and faster and faster. And that's why we describe it as a as a muscle rather than as a transformation that at some point we will be done.

[00:10:36]So, Keith, one of the most asked questions we got from our audience ahead of this session was how to get started.

[00:10:44]So, for leaders watching who want to be on the right side of that line, what's the most practical approach in your view to next steps?

[00:10:54]Yeah, I love it. And one of the things that we actually did in this second edition is we did what I'll affectionately call the the TLDR version of the book, right? Cuz it's a big book.

[00:11:05]It's got a heck of a third factor to it.

[00:11:07]And we wanted to distill down, look, when you step back from all of this, what are those signature moves, those things that really make a difference here? And therefore, how do you get started against those? I'll pull out maybe a couple of these if I if I may, Lisa, that are maybe personal personal favorites. The first is this this domain change, right? Domains, not individual use cases or end-to-end workflows. But the reason this matters is you are pointing your resource at the points of greatest economic leverage for your industry, right? So, you're actually in essence solving business problems that actually matter and will move the needle for you. And it is striking to me how many companies I I spend time with who, you know, sort of think they did that at the start and they all agreed and stacked hands, but when it really comes down to it, they sort of peanut buttered the resource amongst everybody because everybody had their own version of it.

[00:11:58]So, so, domains not use cases and really pointed at three or fewer that that that move the needle for your company. So, you know, if I'm in retail, that's probably forecasting and planning, right? If I'm in insurance, it's probably claims processing.

[00:12:13]You know, you you know, if I'm in heavy manufacturing, it might be yield or or throughput. Another one that I personally love is is the piece on on talent density, right? So, if you look at number five here, every AI transformation is at its heart a people transformation. We wrote that in in book one. That is more true today than it has ever been.

[00:12:33]And the reason, of course, is the level of change that is happening around us, right? The talent density that you have on your teams matters, particularly the technical talent density.

[00:12:43]And at the same time, as we move into, you know, agents taking more of the kind of coordination type roles, more of the routine execution and decision-making, then then actually the human role shifts up that value stack as well. And starting to think about what having, you know, both carbon and silicon employees together in one organization, how that works, the level of people change around that becomes really consequential. So, if you're not thinking about your AI transformation as a set of people changes, then you're probably off track somewhere in what you're you're doing.

[00:13:15]So, those are those are two two favorites of mine. I won't go through all 12. I don't know, Rob, if you have a a particular favorite as well. You know, speed is a defining organizational advantage is interesting. You you mentioned DBS, and I think that the $1 billion of realized value is impressive.

[00:13:29]The other North Star metric I love about DBS is, you know, when they started out with AI, it took them 18 months to get their first model to production. Now they put a model into production every 2 months, right? That speed, that that time like that is that's the capability.

[00:13:44]That's that's the differentiated capability that allows them to just keep going um doing more and going faster. Uh so I think that's interesting. Um and then I think probably, you know, uh you know I'm a fan of agentic engineering. Uh you know, I think this is the fundamental disruption that we're only beginning to see the potential of.

[00:14:03]Uh the great example in the book actually is LATAM Airlines, uh which, you know, um I'm not sure all of us would would sort of um think about airlines as, you know, the bleeding edge of technology, but LATAM Airlines has is probably a year ahead of most companies in terms of fully adopting and embedding agentic engineering, not just for coding, but for the entire software development life cycle.

[00:14:25]And they're going so fast as a result.

[00:14:27]And also related to this point, it it's changed the talent mix, right?

[00:14:31]We really need uh you know, great engineers, but we need those who have retooled themselves with agentic software. Amazing. Thank you both. Okay, these points converge on on a single unavoidable question, which is who owns this?

[00:14:48]A transformation this consequential, touching every capability, every domain, every layer of the business, even the most techno-optimistic among us have to concede that it's not going to run itself. So, Kate, where does accountability sit here?

[00:15:05]Yeah.

[00:15:06]Um it's both top-down and it's distributed, right? So, in all the hundreds of transformations that we've now now studied in our in our misspent youth, um there isn't a single one that is successful that does not have this as a number one, number two priority by the CEO, right? That's a given. I've not found it.

[00:15:24]Um at the same time, the ownership of how to actually do this has to be distributed across the the full leadership team. So, I describe it as it's it's a corporate team sport, right?

[00:15:34]Um it's actually got to be everybody's job. And one of my telltale signs, you know, walking into a uh you know, management team or whatever, is when someone asks that question in the room of about technology, let's say, and you watch everybody in the room sort of turn and face the one person that's got it in their job title, right? That's when you know that this is not going to work because you need your CHRO to wake up in the in the morning and say, "Actually, what is an agentic organization going to look like?" Your CFO is, you know, rewiring the funding mechanisms that are going to allow you to actually reinvest and and uh you know, invest and reinvest in this over time. Your your business owners, you know, the domain owners, the real heart of the transformation, uh you know, need to actually, you know, own this, right?

[00:16:18]So, it is both top-down and it is uh distributed as well. I think longer the days now where you can delegate this to uh the technology function and and and hope for a for a for a good outcome.

[00:16:28]It's just not enough anymore.

[00:16:30]Okay.

[00:16:31]Thanks, both. I am mindful of time and we've been getting a slew of questions.

[00:16:36]I'd like to turn to Q&A, at least try to squeeze in one or two. Rob, what are the gaps most businesses miss in their AI maturity journey? One of the first missteps is is kind of I think this mindset that the job of an ELT is to listen to proposals around AI, resource them, and then turn to sort of the CDIO uh capability to sort of get it done.

[00:17:01]And we would say, "No, it's it's if you if the if your mindset is I I run a call center or I run supply chain planning, and I'm going to totally reinvent how this is done." I think, you know, you you're it's much more than a technology thought, right? It needs to be entirely business-led. So, that's one thing I think folks fall back on an old paradigm of working with IT and that just doesn't work in this day and age.

[00:17:23]Um that's one I think obstacle.

[00:17:26]I think another obstacle we see is just is just the lack sort of a lack of preparation for adoption.

[00:17:33]In at least two two way two three ways. One is, you know, very often we we resource things to MVP. We have an idea, let's go invest the money and see if we can prove it out. I think we're less interested in the number of successful MVPs. We're more interested in how many of these end-to-end workflows have fully scaled across the entire footprint of your of a company where it's relevant. Um and so from adoption, I have you resourced that?

[00:17:56]Have we thought about the fact when the MVP works, we then need to get this to production and rinse and repeat it and scale it across our business.

[00:18:04]Have we thought about the fact that as we rinse and repeat that, we need to an efficient way of, you know, kidding this technology so that when I bring it to the next country or the next product line, I don't need to fully reinvent the wheel. I'm just sort of tailoring from there. Um have we thought about the adoption that is not technical? We like the example of a big automotive company in the book that fully reinvented their supply chain.

[00:18:26]But the harder as hard as that was, they then had to work with hundreds of suppliers to sort of get them to work in the way that, you know, the digital twin of the supply chain said it should.

[00:18:35]That's incredible change. These aspects of adoption are things that they're often afterthoughts and therefore they get become stumbling blocks and slow our progress.

[00:18:44]Okay, that's super helpful. I'm going to squeeze in one more. Kate, very quickly, what does it take to build AI conviction both yourself as a leader and across your organization before that window to lead actually closes?

[00:18:58]Yeah. Let me say two very quick things maybe on this one. I love the question.

[00:19:03]The first is the best way to build conviction is is focus on the value, right? So, follow where the money actually is and solve real business problems. You want to build conviction, You know, that's the that's the heart.

[00:19:14]And the second thing I would say is also cut yourself a little bit of slack on this. This stuff is hard. And one of the maybe, you know, benefits of of agenic and so on is that the cost of iteration has come down, right? If you like the cost of a wrong turn has got or it's easier to make a wrong turn and and pivot. And, you know, I think it's becoming less and less about generating the or coming up with the perfect answer, much more about you know, stress testing it, owning it, building the conviction around it, building the, you know, the right to to deliver that that change. And I think there's beauty in the messiness of the process sometimes and what ends up on the cutting room floor and why maybe where the real value value sits.

[00:19:57]So, I would cut some slack as well.

[00:19:59]Inspiring. Okay, that brings us to time.

[00:20:01]Fantastic discussion, Robin Keith.

[00:20:03]Thanks for joining us today and again, big congratulations on the launch of Rewired.

[00:20:09]Thank you so much.

[00:20:11]>> [music] >> Thanks so much for listening to the McKinsey podcast. I'm Lucia Rahilly. And I'm Roberta Fasaro. Find us on mckinsey.com. [music] We'll have a transcript of this episode up shortly. And download the McKinsey Insights app where you can find this podcast and other helpful content [music] updated daily. If you enjoyed the show, we'd love for you to leave a rating [music] and a review. We'll see you next time.

[00:20:38]>> [music]